Els dies intensos de calor de l’estiu han portat algun article interessant d’intel·ligència artificial. Per exemple, en un article al JCTC esmentat al C&EN: Machine learning predicts electronic properties at relatively low computational cost (article: Transferability in Machine Learning for Electronic Structure via the Molecular Orbital Basis).

The researchers wanted to predict electronic structure correlation energies—a measure of interactions between electrons that helps chemists model how a molecule behaves. Their machine-learning approach predicts these values based on a set of known data.

Miller’s group trained its algorithm on localized molecular orbitals of a set of small molecules. Because molecular orbitals are agnostic to the underlying bonds and atoms, Miller says the new algorithm could predict properties for many different molecules with a small starting set of data.

In one example, the researchers trained their algorithm on the molecular orbitals of water, then predicted the correlation energies of ammonia, methane, and hydrogen fluoride. For methane, the algorithm’s value was just 0.24% off from the one generated by CC, and that was the least accurate result of the three they found. The algorithm’s calculation for a cluster of six water molecules took two minutes with machine learning, compared with 28 hours for CC.

I de l’abstract del paper:

We present a machine learning (ML) method for predicting electronic structure correlation energies using Hartree–Fock input. The total correlation energy is expressed in terms of individual and pair contributions from occupied molecular orbitals, and Gaussian process regression is used to predict these contributions from a feature set that is based on molecular orbital properties, such as Fock, Coulomb, and exchange matrix elements. With the aim of maximizing transferability across chemical systems and compactness of the feature set, we avoid the usual specification of ML features in terms of atom- or geometry-specific information, such atom/element-types, bond-types, or local molecular structure. ML predictions of MP2 and CCSD energies are presented for a range of systems, demonstrating that the method maintains accuracy while providing transferability both within and across chemical families; this includes predictions for molecules with atom-types and elements that are not included in the training set. The method holds promise both in its current form and as a proof-of-principle for the use of ML in the design of generalized density-matrix functionals.

Serà interessant veure com progressa aquest tema!

I al Chemistry World, un article sobre nous fàrmacs: Artificial intelligence seeks out new anticancer drugs. És un paywall, així que no puc analitzar l’article, però em ve ara al cap que a l’IQCCUdG durant anys el grup d’en Ramon Carbó-Dorca ha fet prediccions d’estructura-activitat fent servir QSAR. Caldrà veure l’article original.

Un altre de Chemistry World: AI robot tests, predicts and even discovers reactions that are new to chemistry. Aquí sembla que es tracta d’un robot real, articular, que treballa d’una forma molt fina al laboratori experimental.

Ens trobem doncs amb un parell de casos de dificultat d’accés a una revista secundària, pensada per difondre i divulgar la recerca. Ens queixem força que hi ha pocs articles Open Access… però i aquests articles de Chemistry World i altres revistes que només són accesibles si es paga, és a dir, que són darrere un paywall?

News Reporter